Hello, Just curious whether it would be possible to convert an [Open]Solaris or Nevada installation onto a physical server into an xVM domain? If I create and HVM Domain which points to the whole physical disk where the installation was done, would that work? or is the hypervisor/Domain-0 expecting to use the disk differently? I assume I would have to muck around with the device names since the disk and network devices would change running as a Domain. --joe
Joseph Mocker wrote:> Hello, > > Just curious whether it would be possible to convert an [Open]Solaris or > Nevada installation onto a physical server into an xVM domain? > > If I create and HVM Domain which points to the whole physical disk where > the installation was done, would that work? or is the > hypervisor/Domain-0 expecting to use the disk differently? > > I assume I would have to muck around with the device names since the > disk and network devices would change running as a Domain.I''ve not heard of anyone trying this yet, but yes, it should be possible and should be fairly straightforward. You''ll at the least have to fix up the boot device path, which you should be able to do from failsafe boot. Before you try with the physical disk, dd it to a file and try booting from that. Another thing to try is to boot it as a PV domain. ----------------------------------------------------- Russ Blaine | Solaris Kernel | russell.blaine@sun.com
Yes, this should be quite doable. The procedure should the same as is used quite commonly to move from 1 hardware platform to another. This is the procedure I used, it was posted on the Yahoo x86 group a long time back. I don''t remember who it was, so can''t give appropriate credit... Procedure: 1. Take a backup. Rince and repeat. Take another backup as well. 2. change hardware 3. Boot failsafe single user on new hardware - normally your root disk gets mounted as /a doing this. 4. Replace /a/etc/path_to_inst with the failsafe /etc/path_to_inst (should be a single stub line) 5. Remove the physical devices from /a/devices i.e the pci@* paths (rm -rf). Dont touch any pseudo or others you are unsure of. 6. do an rm /dev/dsk/*, /dev/rdsk/* to get rid of the old entries 7. run devfsadm to fix up the device config from scratch: # devfsadm -r /a 8. Check that the entries in /a/devices and /a/etc/path_to_inst are now populated. If not, use tar to copy the failsafe device tree in place and run step 7 again i.e # cd /devices; tar cvf -|(cd /a/devices; tar xf -) 9. Check /a/dev/dsk - make sure you have disks. 10. edit /a/boot/solaris/bootenv.rc - fix your boot-device path 11. Check no disk names have changed - if they have, fix /etc/vfstab 12. Run "bootadm update-archive -R /a" to repack your boot archive. You should now be able to boot off the disk as if it always lived on your system. this is really just a modification of an old procedure for SPARC systems to re-order controller numbers. There is probably an Infodoc on it somewhere. -- This message posted from opensolaris.org